A detective enters a rain-soaked city searching for a vampire.
What begins as a simple hunt leads him beneath the streets, where he discovers that the vampire he has been chasing may not be the real enemy.
Detective vs Vampire is an experimental AI animated noir-horror short created scene by scene using MiniMax H3 running locally on an NVIDIA DGX Spark.
For this project, I tested whether MiniMax H3 could maintain a continuous story across multiple generations while handling:
• Recurring characters
• Dialogue and lip synchronization
• Cinematic camera movement
• Rain and environmental effects
• Continuous noir-style music
• Sound effects and ambience
• Action and fight scenes
• Multiple locations
• Character interactions
• A beginning, story progression and ending
Each scene was generated separately. I extracted the final frame of each clip and used it as the starting frame for the next generation, allowing the story to continue from scene to scene.
The video, character voices, music, sound effects and environmental audio were generated with MiniMax H3.
This is still experimental AI filmmaking, so you may notice some visual or continuity imperfections—but seeing how far a locally running video model can take a continuous story is what made this experiment so interesting.
Headphones recommended for the full noir soundtrack and atmosphere.
Subscribe for more locally generated AI films and DGX Spark experiments.
#AIShortFilm #MiniMaxH3 #AIMovie #AIAnimation #DetectiveStory #Vampire #Noir #HorrorAnimation #AIFilm #GenerativeAI #DGXSpark #LocalAI #ComfyUI
ข้อเสนอแนะ
รายงาน
ห้ามทำซ้ำหรือดัดแปลงโดยไม่ได้รับอนุญาตจากครีเอเตอร์
ข้อเสนอแนะ
รายงาน